Commit Is a Save Point!

• “The purpose of Git” is to create a COMMIT
• A commit is a save point of a state of the project folder

You Can Go Back to Any Commit

You Can Compare Any Two Commits

What Is Commit in a Computer?

• If you init a git project on a folder, a hidden folder .git will be made
• When you create a commit, the state of the folder will be stored in .git folder
• The name of commit is just a random hash string

Local & Remote Repository

• .git folder in your local machine (like in the previous page) is called local repository
• Usually, a remote repository (like GitHub) is used for code sharing
• Push and Pull is a command to sync your local repository with the remote repository

Stage

• When you modify some files, you can create a commit only for specific changes
• This choosing process is called, Staging

Command Line

git add foo1.txt

VS Code

Commit

• You have to add a message for each commit

Command Line

git commit -m "YOUR COMMENT"

VS Code

Push and Pull

For now, forget about origin main

git push origin main

git push origin main

Handson 1: First Commit and Push

1. Go to GitHub webpage and create a repository

1. Choose the private repository (if it is for research)
2. Choose the language you are using for .gitignore
3. Add README
4. No need to choose license (because it is private)

2. Clone it to your local machine

1. The easist way is to open a new window of VS Code and choose “clone a repository”
2. You can also do git clone URL_OF_REPOSITORY

3. Create a Commit and Push it

1. Create a text file (say foo1.txt) with a random text (say “Hello Git!”)
2. Create a commit (don’t forget to stage it before)
3. Push to the remote repository

HEAD and Branch

• Branch: a label of a commit
• HEAD: the branch you are seeing

• To create a branch, git branch BRANCH_NAME
• To move the HEAD, git switch BRANCH_NAME

Merge

• You can merge a branche into HEAD by git merge BRANCH_NAME

1. Merge at Local Repository

• Merge it to main on your local repository
• Push the local main into the remote main
• You can use it for solo project

2. Merge at Remote Repository

• Create a Pull Request on GitHub Repository
• Decide to merge it into main or not on GitHub
• Used for collaboration (but I use it even for solo project)

Questions about Branches

Why Do We Use Branch?

• Keep main branch clean and work correctly
• Easy to detect a bug (because “main” works perfectly)
• Collaborators create a branch from main and work on it

When Should I Create a Branch?

A simple suggestion is main-dev workflow

• You only work on dev branch (with a collaborator, I use my name for the developing branch)
• When you finish a meaningful chunk of works, cleanup uneccessary files and merge dev into main

I usually merge dev into main just before the meeting with my supervisor or collaborators.

Conflict

A different code is written in the same line of two branches in merging

aaaaa
bbbbb
<<<<<<< HEAD
cccccc
======
zzzzzz
>>>>>>> dev
ddddd

Solve a Conflict

1. Delete the lines you don’t need and save the file
2. Create a commit for this merge and solving the conflict

aaaaa
bbbbb
<<<<<<< HEAD
cccccc
======
zzzzzz
>>>>>>> dev
ddddd

aaaaa
bbbbb
zzzzz
ddddd

Pull = Fetch + Merge

• Fetch : download a branch from the remote repository (origin) to a local machine, named origin/BRANCH_NAME
• git pull origin main is equivalent to git fetch origin main && git merge origin/main
• Because it is a merge, git pull could return a CONFLICT!

Handson 2: Two Types of Merge

1. Local Merge

1. Create a new branch “dev” and switch the HEAD into it
2. Create some commits
3. Switch HEAD into main
4. Merge dev into main
5. Push it to origin/main
6. Delete (local) dev branch

2. Remote Merge

1. Create dev branch and switch into it
2. Create some commits
3. Push it to origin/dev
4. Go to GitHub and create a Pull Request
5. Merge it into main and delete the remote dev branch
6. Go back to local and switch into main
7. Pull main branch and delete (local) dev branch

Handson 3: Three Types of Conflicts

1. Local Conflict

1. Switch to dev and create foo1.txt with a line of text (e.g. “Tortilla sin Cebolla”) and create a commit
2. Switch to main and create foo1.txt with a different line of text (e.g. “Tortilla con Cebolla”) and a create a commit
3. Merge dev into main, which produces a conflict
4. Solve the conflict by choosing either of the lines

2. Pull Request Conflict

1. Head on main and create foo2txt with a line of text (e.g. “Oh! How beautiful are our Emperor’s new clothes!”). Push it into origin/main
2. Switch to dev and create foo2.txt with a different line of text (e.g. “But the Emperor has nothing at all on!”). Push it into origin/dev
3. Go to GitHub and create a pull request. Solve the conflict on GitHub

3. Pull Conflict

1. Switch to dev and create foo3.txt with a line of text (e.g. “Mountain of Mushroom”), create a commit, and push into origin/dev
2. Go to GitHub and merge dev into main
3. Go back to local. Switch to main and create foo3.txt with a different line of text (e.g. “Mountain of Bumbooshoot”), create a commit
4. Pull from origin/main, which produces a conflict. Solve it.

1. Sync Local Repository

git switch main
git pull origin main

2. Write Your Codes on Dev branch

1. Switch to a new developing branch

git branch dev
git switch dev

Or

git checkout -b "dev"

2. Write your codes

3. Commit

1. Stage files you want to commit

git add foo1.txt foo2.txt

If you want to stage all modified files, git add .

2. Commit

git commit -m "hogehoge"

4. Push to the Remote Repository

git push origin dev

5. Pull Request and Merge

I recommend you to delete the remote and local dev branch here.

Handson 4: Follow My Workflow

• Follow my workflow 1-5 with a few commits
• Try both command line and VS Code

Data Version Control (DVC)

How Can We Work with Data in Git?

• We want to store the data in the Git project (because referred in the code)
• Not interested in change in each line of the data (imagine data cleaning)
• There is a limit for the file size in GitHub (100 MB)

How Does DVC Work?

• Create a text file for the meta-information of the data
• Git manages only the text file
• Git can follow when the data is added, modified, deleted through the text file
• The original data is stored in the remote storage (Google Drive, Amazon S3, ..)

How Does DVC Work?

Setup Commands

1. Init a DVC project

dvc init

2. Setup the remote DVC Repository (Google Drive)

• Create a new Google drive folder and copy the folder id from its URL
• Run the follwoing

dvc remote add --default myremote gdrive://GDRIVE_FOLDER_CODE

Workflow Commands

1. Track a file by DVC

dvc add foo1.csv

You can specify a folder (e.g. data) with -R option dvc add -R data

2. Push data to the remote storage

dvc push

3. Pull data from the remote storage

dvc pull

Handson 5: Setup DVC and Follow My Workflow

1. Init a DVC project and setup its remote on Google Drive
2. Go to the repository of this workshop and read my workflow
3. In step 2, add a csv file in data/csv folder and create a commit (say, “add foo1.csv”)
4. Follow the rest of my workflow

Troubleshoot

Git Reference log

You can see all the git activity (commit, merge, rebase, …)

git reflog

Git Reset

You can reset any git activity

git reset --soft COMMIT_ID_OR_REFLOG_ID

Keep in Mind

• Nothing bad happens with Git. It just stores savepoints.
• In practice, you rarely restore a file from a commit. But it gives you a lot of relief and freedom.

Learn More

Introduction to Git・ Introduction to GitHub